Motion Retargetting based on Dilated Convolutions and Skeleton-specific Loss Functions

Sang Bin Kim; Inbum Park; Seongsu Kwon; Jung Hyun Han

doi:10.1111/cgf.13947

Motion Retargetting based on Dilated Convolutions and Skeleton-specific Loss Functions

Sang Bin Kim, Inbum Park, Seongsu Kwon, Jung Hyun Han

Department of Computer Science and Engineering

Research output: Contribution to journal › Article › peer-review

5 Citations (Scopus)

Abstract

Motion retargetting refers to the process of adapting the motion of a source character to a target. This paper presents a motion retargetting model based on temporal dilated convolutions. In an unsupervised manner, the model generates realistic motions for various humanoid characters. The retargetted motions not only preserve the high-frequency detail of the input motions but also produce natural and stable trajectories despite the skeleton size differences between the source and target. Extensive experiments are made using a 3D character motion dataset and a motion capture dataset. Both qualitative and quantitative comparisons against prior methods demonstrate the effectiveness and robustness of our method.

Original language	English
Pages (from-to)	497-507
Number of pages	11
Journal	Computer Graphics Forum
Volume	39
Issue number	2
DOIs	https://doi.org/10.1111/cgf.13947
Publication status	Published - 2020 May 1

Bibliographical note

Funding Information:
This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korea government (MSIT) (NRF‐2017M3C4A7066316 and No. NRF2016‐R1A2B3014319).

Funding Information:
This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korea government (MSIT) (NRF-2017M3C4A7066316 and No. NRF2016-R1A2B3014319).

Publisher Copyright:
© 2020 The Author(s) Computer Graphics Forum © 2020 The Eurographics Association and John Wiley & Sons Ltd. Published by John Wiley & Sons Ltd.

Keywords

CCS Concepts
• Computing methodologies → Neural networks

ASJC Scopus subject areas

Computer Graphics and Computer-Aided Design

Access to Document

10.1111/cgf.13947

Cite this

@article{bb2955b41d7e4d2f931c1a0ce8d8e6b3,

title = "Motion Retargetting based on Dilated Convolutions and Skeleton-specific Loss Functions",

abstract = "Motion retargetting refers to the process of adapting the motion of a source character to a target. This paper presents a motion retargetting model based on temporal dilated convolutions. In an unsupervised manner, the model generates realistic motions for various humanoid characters. The retargetted motions not only preserve the high-frequency detail of the input motions but also produce natural and stable trajectories despite the skeleton size differences between the source and target. Extensive experiments are made using a 3D character motion dataset and a motion capture dataset. Both qualitative and quantitative comparisons against prior methods demonstrate the effectiveness and robustness of our method.",

keywords = "CCS Concepts, • Computing methodologies → Neural networks",

author = "Kim, {Sang Bin} and Inbum Park and Seongsu Kwon and Han, {Jung Hyun}",

note = "Funding Information: This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korea government (MSIT) (NRF‐2017M3C4A7066316 and No. NRF2016‐R1A2B3014319). Funding Information: This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korea government (MSIT) (NRF-2017M3C4A7066316 and No. NRF2016-R1A2B3014319). Publisher Copyright: {\textcopyright} 2020 The Author(s) Computer Graphics Forum {\textcopyright} 2020 The Eurographics Association and John Wiley & Sons Ltd. Published by John Wiley & Sons Ltd.",

year = "2020",

month = may,

day = "1",

doi = "10.1111/cgf.13947",

language = "English",

volume = "39",

pages = "497--507",

journal = "Computer Graphics Forum",

issn = "0167-7055",

publisher = "Wiley-Blackwell",

number = "2",

}

TY - JOUR

T1 - Motion Retargetting based on Dilated Convolutions and Skeleton-specific Loss Functions

AU - Kim, Sang Bin

AU - Park, Inbum

AU - Kwon, Seongsu

AU - Han, Jung Hyun

N1 - Funding Information: This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korea government (MSIT) (NRF‐2017M3C4A7066316 and No. NRF2016‐R1A2B3014319). Funding Information: This work was supported by the National Research Foundation of Korea (NRF) Grant funded by the Korea government (MSIT) (NRF-2017M3C4A7066316 and No. NRF2016-R1A2B3014319). Publisher Copyright: © 2020 The Author(s) Computer Graphics Forum © 2020 The Eurographics Association and John Wiley & Sons Ltd. Published by John Wiley & Sons Ltd.

PY - 2020/5/1

Y1 - 2020/5/1

N2 - Motion retargetting refers to the process of adapting the motion of a source character to a target. This paper presents a motion retargetting model based on temporal dilated convolutions. In an unsupervised manner, the model generates realistic motions for various humanoid characters. The retargetted motions not only preserve the high-frequency detail of the input motions but also produce natural and stable trajectories despite the skeleton size differences between the source and target. Extensive experiments are made using a 3D character motion dataset and a motion capture dataset. Both qualitative and quantitative comparisons against prior methods demonstrate the effectiveness and robustness of our method.

AB - Motion retargetting refers to the process of adapting the motion of a source character to a target. This paper presents a motion retargetting model based on temporal dilated convolutions. In an unsupervised manner, the model generates realistic motions for various humanoid characters. The retargetted motions not only preserve the high-frequency detail of the input motions but also produce natural and stable trajectories despite the skeleton size differences between the source and target. Extensive experiments are made using a 3D character motion dataset and a motion capture dataset. Both qualitative and quantitative comparisons against prior methods demonstrate the effectiveness and robustness of our method.

KW - CCS Concepts

KW - • Computing methodologies → Neural networks

UR - http://www.scopus.com/inward/record.url?scp=85087780809&partnerID=8YFLogxK

U2 - 10.1111/cgf.13947

DO - 10.1111/cgf.13947

M3 - Article

AN - SCOPUS:85087780809

SN - 0167-7055

VL - 39

SP - 497

EP - 507

JO - Computer Graphics Forum

JF - Computer Graphics Forum

IS - 2

ER -

Motion Retargetting based on Dilated Convolutions and Skeleton-specific Loss Functions

Abstract

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Cite this